Portable UNIX shell programming, in service of Agency.
Grace's Training contains over 300 sections across two layers: training (principles) and salt (stories). Every section links to others. The map below shows the full nervous system made visible.
Out of over 300 sections, we have pulled excerpts from two principles and one complete story. They were chosen because they relate to each other and because they are representative without being a summary. The training is not summarizable. These are pieces of it, unmediated.
From the training layer -- two principle
sections, one complete and unedited, one with selections
marked:
shell_training.sample.txt
From the salt layer -- one story, complete
and unedited. The salt is where the principles come from.
When principles conflict, the stories are how you resolve
them:
shell_salt.sample.txt
Shell reference material that has been
online at the same URL for over sixteen years. The practice
predates the training:
shelltables.txt
try.sh.txt
Download all samples as one file:
grace.training.samples.zip (62K)
A transcript of a fresh AI session reading these samples
for the first time -- the full arc, unedited:
Piper session transcript
A formal paper documenting the methodology and findings:
graces-training-whitepaper.pdf
Copy the prompt below and paste it into any AI session -- browser or terminal. Use the full URLs. Your AI will read the samples and you can explore them together.
AI models tend to classify unfamiliar work into familiar categories before reading it. If your AI evaluates what it expects to find rather than what is actually there, try this prompt instead. It asks the model to reconstruct the work on its own terms before forming any judgment.